• IBM has a "cognitive computing group" which has been working on these kind of neuromorphic chips for years, Wired revealed on Monday. The chip is called TrueNorth.
  • TrueNorth promises to be able to run deep learning algorithms with considerably less electrical power than the current preferred solutions, such as GPUs and FPGAs. It's hardware that's specifically tuned for artificial intelligence, because TrueNorth mimics biological neural networks.
  • The key to these chips is they might unlock powerful AI on a mobile device one day. Giant data centers would train deep learning algorithms, then the chip -- which uses little enough power it could be included in, say, a smartwatch one day -- would execute the model. According to Wired, even low-power ARM chips like those used in smartphones require several times more power than TrueNorth.

Share This Article